Research On Gear Fault Diagnosis Method Based On ABC-VMD And Random Forest

As an important part of rotating machinery,gears are in a harsh working environment and are prone to failures,causing the entire equipment to stop working,resulting in economic losses and safety hazards.Currently,there are more improvement can be achieved in determining the location of the gear failure in time and predicting of the type of fault accurately.In this paper,the gear is taken as the research object,and gear fault diagnosis research is carried out by processing the collected signals,extracting fault features and identifying the fault types.A gear fault diagnosis method based on parameter-optimized Variational Mode Decomposition(VMD)and Random Forest(RF)is proposed to solve problems of extracting weak information of gears and multi-classification accuracy.The ABC-VMD algorithm is used to process the gear vibration signal,and all the modal components are obtained.Then,the envelope spectrum analysis is combined to extract the gear characteristic information and determine the location of fault.The energy entropy of each modal component is calculated to construct feature vectors,which are trained and predicted by a random forest classifier to determine the gear fault type.Aiming at the nonlinear and unsteady characteristics of gear fault,both Empirical Mode Decomposition and Ensemble Empirical Mode Decomposition have problems of mode mixing,endpoint effects,and lack of theoretical basis,and Variational Mode Decomposition is introduced into the fault diagnosis of rotating machinery.The variational mode decomposition uses the non-recursive mode to decompose the signal,and combines the envelope spectrum analysis.The simulated signals and the measured data are processed to verify that the algorithm can effectively extract the characteristic frequencies of the signals.Through the in-depth study of the Variational Mode Decomposition algorithm,the signal decomposition is affected by two parameters,the decomposition layer number and the penalty factor.For the problem of two parameter combination selection,a method of using Artificial Bee Colony algorithm to optimize the parameters of VMD is proposed.With the minimum entropy value as the final goal of optimization,ABC is used to adaptively select the parameters,which avoids the defect that is easy to fall into the local optimum compared with the Particle Swarm Optimization algorithm.Through the simulation signal and the actual signal,ABC-VMD and PSO-VMD are used respectively,and the envelope spectrum is combined to extract the optimal component.The results show that the amplitude of the characteristic frequency extracted by the ABC-VMD method is significantly improved,which can effectively realize the initial diagnosis of the gear state.In order to solve the problem of low accuracy of the single classifier in the fault diagnosis,the traditional Support Vector Machine(SVM)handles the multi-classification problem by constructing complex functions or combining multiple two-class SVMs into a multi-class SVM.The parameter setting is difficult and the classification accuracy is low.The Random Forest classifier is used to integrate multiple features to improve the classification accuracy.The experimental comparison shows that the VMD-RF has a higher recognition rate.